01
Let the pair take over
Choose The corridor. Gain is high and intention modest. The two fall into a rhythm several times faster than either set out to keep, and the faint lines behind them show what each was aiming for.
communicoupling
Two people step aside
as one.
Two people meet in a corridor. One steps left; the other mirrors it. Both step back and mirror again. Every move is meant to end the encounter and every move extends it. Neither of them wrote the pattern that now holds them there.
Hanne De Jaegher and Ezequiel Di Paolo call this participatory sense-making. Meaning gets generated inside the interaction and belongs to neither participant: the conversation neither person could have had alone, the joke that exists between exactly these two. The stronger claim is that the interaction process acquires a life of its own and resists both participants' intentions.
individual intentionhow firmly each agent holds its own course
coordination gainhow strongly each adjusts to the other's last move
noisespontaneous variation neither agent planned
emergent pattern
—
share of the pair's motion supplied by the coupling
lock episodes
—
stretches that hold steady while both agents are off their own course
both goals met
—
share of the run with each agent on the course it intended
One encounter, two coupled agents
gold and blue: each agent's side-to-side position · faint lines: the course each intended alone · violet: stretches the pair sustains against both
Set the encounter
what to try
Find out whose pattern it is.
02
Let coordination serve
Choose The duet. Moderate gain works against a small difference in tempo. The pair holds a steady relation while each agent stays on its own course.
03
Watch alignment mislead
Choose The interrogation. Alignment climbs to its ceiling while one agent holds its course and the other is carried onto it. Lock episodes fall to none.
the mechanism
A lagged response answers a moment that has passed.
Each agent carries an intended course, a phase advancing at the rate it means to keep, and pulls itself toward that course with its intention strength. Each also pulls toward where it last saw the other, a dozen steps back. The lag does the work. Because every response arrives late, mutual adjustment stops settling and starts driving, and above a threshold in gain the two leave both intended courses for a rhythm that belongs to the coupling.
A step counts as locked when the relation between the agents holds steady while both sit away from their own courses. Alignment on its own would miss this. Two agents with identical habits and no coupling at all look perfectly aligned, so the model measures the joint pattern's grip separately from how synchronized the pair appears.
the social payoff
Some patterns cannot be stopped one participant at a time.
The argument that restarts whenever the topic comes near it, the meeting that finds an agenda nobody tabled, the two colleagues who cannot hold a short conversation: each is a candidate for a pattern with its own momentum. The diagnosis carries a prediction. An intervention aimed at one participant's intentions leaves the pattern standing, and the participant who tries hardest to end it may be feeding it.
This places participatory sense-making downstream of double contingency and alongside gesture and response. The coupling that makes shared meaning possible keeps producing it after both parties would rather stop.
the mapping
Mechanism ↔ social life.
| In the model | In the world |
|---|---|
| individual intention | How firmly each party holds to what it came to do. |
| coordination gain | How strongly each adjusts to the other's last observable move. |
| reaction lag | Responses land after the moment they answer. Held fixed at a dozen steps. |
| noise | Spontaneous variation. Large enough, it knocks the pair out of a lock. |
| emergent pattern | Motion supplied by the coupling term as a share of all deterministic drive. |
| lock episode | A stretch where the pair's relation holds steady while both agents sit away from their own courses. |
| alignment | How constant the relation between the two stays across the run. |
where it tears
Limits.
The corridor is the strong case and close to the only clean one.
Most interactions stay individually steerable most of the time. One party changes tack and the pattern follows. Treating every conversation as an agent in its own right re-mystifies the thing the concept was built to explain.
"The interaction did it" can excuse the powerful participant.
An interrogation is not a duet. The interrogation preset shows the failure mode: alignment at its ceiling, one course intact, the other abandoned, and no episode that holds against both parties at once. Asymmetry of intention strength is where the concept needs the most care.
Process autonomy is hard to demonstrate outside toy cases.
The lock criterion here is a stipulation: a threshold on how far an agent has drifted from its own course, a window over which the pair's relation must stay steady. Move either and the count moves. Whether real interaction shows autonomy in the strong sense remains contested.